OpenCode
OpenCode is an open-source AI coding agent that gives developers a terminal-first agent surface with installable CLI builds, desktop beta support, plan/build agent modes, and subagent workflows around cloud or local model choices.
OpenCode is one of the strongest HN-backed examples of an open coding-agent product. It helps readers compare Claude Code, Codex CLI, Cursor-style workflows, and local coding agents through the harness around the model: install surface, plan mode, shell execution, review loops, and model routing.
The official GitHub repository describes OpenCode as an open-source coding agent, links to opencode.ai, documents CLI installation through script and package managers, lists Windows/macOS/Linux desktop beta downloads, and describes built-in build and plan agents plus a general subagent. Hacker News demand evidence around "OpenCode - Open source AI coding agent" shows strong developer discussion, but HN should be used for demand and comparison language rather than factual claims.
- Run an open-source coding agent from a terminal-first workflow.
- Use plan mode for read-only exploration before granting edit or shell permissions.
- Compare open coding-agent harnesses against Claude Code, Codex CLI, Cursor, Aider, and local-model agents.
- Evaluate whether desktop, CLI, package-manager, and subagent surfaces fit a team workflow.
OpenCode is maintained as an open-source coding-agent repository with MIT licensing and official documentation. Its README documents installation through curl, npm, Scoop, Chocolatey, Homebrew, Arch, mise, and Nix, plus desktop beta downloads for macOS, Windows, and Linux.
- Agent modes: a build agent for development work and a plan agent that is read-only by default.
- Subagent support: a general subagent can handle complex searches and multistep tasks.
- Surfaces: CLI/TUI is the main source of truth, with desktop builds available in beta.
OpenCode is useful on GetLLMs because it shows that a coding agent is more than the model endpoint. Plan mode, permissions, shell execution, install targets, desktop wrappers, and subagents all shape whether the agent feels inspectable or risky. That makes it a concrete related entity for Agent Harness and Local Coding Agents.
OpenCode has several similarly named websites and community writeups. For page facts, prefer the anomalyco GitHub repository and official OpenCode documentation. Use Hacker News and Reddit only to understand comparison intent, adoption friction, and why readers compare OpenCode with Claude Code, Codex CLI, Cursor, Aider, and local models.
The non-model layer that makes OpenCode-style coding agents useful and controllable.
The category where OpenCode overlaps with local and self-hosted model routing.
The production-readiness question that open coding-agent workflows still need to answer.
OpenCode FAQ
Page-level questions for OpenCode.
Is OpenCode the same kind of tool as Claude Code?+
OpenCode and Claude Code are both coding-agent tools, but they differ in ecosystem and control surface. OpenCode is an open-source coding agent with terminal-first installation paths and plan/build modes, while Claude Code is Anthropic's hosted-agent product with Anthropic account, model, permission, hook, and enterprise controls.
Why does OpenCode matter for local coding agents?+
OpenCode matters because it makes the harness portable across terminal, package-manager, desktop, and model-routing choices. Even when a model is local or self-hosted, the agent still needs planning, permissions, subagents, shell access, and review loops.